Feature based cross-component sample offset optimization and signaling improvement
In some examples, an apparatus for video decoding includes receiving circuitry and processing circuitry. The processing circuitry receives, from a coded video bitstream, coded information associated with one or more blocks in a current picture, the coded information indicates that a cross-component sample offset (CCSO) filter is used for decoding the one or more blocks. The processing circuitry determines one or more features associated with the one or more blocks, the one or more features is used as a context for signaling a value associated with the CCSO filter. The processing circuitry determines at least a signaled value associated with the CCSO filter based on a context model, the context model using the one or more features as a context, and reconstructs samples in the one or more blocks based on at least the signaled value associated with the CCSO filter.
1. A method of video processing in a decoder, comprising:
receiving, from a coded video bitstream, coded information associated with one or more blocks in a current picture, the coded information indicating that a cross-component sample offset (CCSO) filter is used for decoding the one or more blocks, the CCSO filter being associated with a plurality of CCSO filter parameters;
determining one or more features associated with the one or more blocks, the one or more features being used as a context for signaling a value associated with the CCSO filter;
determining, based on the one or more features, a context model that is used for signaling the value of at least a parameter of the CCSO filter in the plurality of CCSO filter parameters, the context model including different probability values associated with candidate values of the parameter of the CCSO filter with the one or more features being used as the context;
determining at least a signaled value associated with the CCSO filter based on the context model; and
reconstructing samples in the one or more blocks based on at least the signaled value associated with the CCSO filter.
2. The method of claim 1 , wherein the context for signaling the value associated with the CCSO filter comprises at least one of whether a specific CCSO filter parameter is signaled, and/or the context used for entropy coding the specific CCSO filter parameter.
3. The method of claim 1 , wherein the one or more features is indicative of texture information in the one or more blocks, the one or more features comprises at least one of
an edge direction for one or more edges in the one or more blocks;
a range of pixel values in the one or more blocks;
a histogram of pixel values in the one or more blocks; and
gradient information of pixels in the one or more blocks.
4. The method of claim 1 , further comprising:
determining the context model based on the one or more features, the context model comprising an assignment of at least a first probability value to a first candidate value of a filter parameter for the CCSO filter and a second probability value to a second candidate value of the filter parameter for the CCSO filter, the first probability value being higher than the second probability value when the first candidate value is associated with the one or more features.
5. The method of claim 4 , further comprising:
determining, based on the context model, a specific filter parameter for the CCSO filter is not signaled.
6. The method of claim 4 , wherein the one or more features comprise an edge direction, the signaled value is indicative of a filter shape, and the determining the context model further comprises at least one of:
when the edge direction is vertical, determining the context model that comprises a highest probability assigned to a vertical filter shape;
when the edge direction is horizontal, determining the context model that comprises a highest probability assigned to a horizontal filter shape; or
when the edge direction is diagonal, determining the context model that comprises a highest probability assigned to a diagonal filter shape.
7. The method of claim 4 , wherein the one or more features comprise a range of pixel values in the one or more blocks, and the context model comprises an assignment of probability values to candidate band numbers.
8. The method of claim 4 , wherein the one or more features comprise a range of pixel values in the one or more blocks, and the context model comprises an assignment of probability values to candidate quantization steps.
9. The method of claim 1 , wherein the determining at least the signaled value associated with the CCSO filter based the context model further comprises:
determining the signaled value indicative of an on/off decision for the CCSO filter based on the context model that uses the one or more features as the context.
10. The method of claim 9 , wherein the on/off decision is of at least one of a frame level, a tile level, and a block level.
11. The method of claim 9 , wherein the one or more features is indicative of a flatness, and the context model comprises an assignment of a higher probability value to a first candidate value of an off decision for the CCSO filter than to a second candidate value of an on decision for the CCSO filter.
12. The method of claim 9 , wherein the one or more blocks comprises a current block and a neighboring block of the current block, and the context model uses a flatness of the neighboring block as context.
13. The method of claim 9 , wherein the one or more features comprise at least one of a first coded syntax value indicative of whether a skip mode being applied, a second coded syntax value indicative of whether intra or inter prediction mode being applied, a third coded syntax value indicative of whether residuals being zero or not, a fourth coded syntax value indicative of an intra prediction mode direction, and a fifth coded syntax value indicative of a coding block partitioning size.
14. A method of video processing in an encoder, comprising:
determining one or more features associated with one or more blocks in a current picture;
determining to use a cross-component sample offset (CCSO) filter having a plurality of CCSO filter parameters for coding the one or more blocks;
determining at least a value of a parameter of the CCSO filter by skipping a check of at least a candidate value of the parameter based on the one or more features;
determining, based on the one or more features, a context model for signaling the value of the parameter, the context model including different probability values associated with candidate values of the parameter with the one or more features being used as a context; and
generating coded information of the one or more blocks based on the context model, the coded information indicating the value of the parameter associated with the CCSO filter according to the context model.
15. The method of claim 14 , wherein the context for signaling the value associated with the CCSO filter comprises at least one of whether a specific CCSO filter parameter is signaled, and/or the context used for entropy coding the specific CCSO filter parameter.
16. The method of claim 14 , wherein the one or more features is indicative of texture information in the one or more blocks, the one or more features comprises at least one of:
an edge direction for one or more edges in the one or more blocks;
a range of pixel values in the one or more blocks;
a histogram of pixel values in the one or more blocks; and
gradient information of pixels in the one or more blocks.
17. The method of claim 14 , wherein the determining the context model comprises:
determining the context model based on the one or more features, the context model comprising an assignment of at least a first probability value to a first candidate value of a filter parameter for the CCSO filter and a second probability value to a second candidate value of the filter parameter for the CCSO filter, the first probability value being higher than the second probability value when the first candidate value is associated with the one or more features.
18. The method of claim 17 , further comprising:
determining, based on the context model, not to signal a specific filter parameter for the CCSO filter.
19. The method of claim 17 , wherein the one or more features comprise an edge direction, the value is indicative of a filter shape, and the determining the context model further comprises at least one of:
when the edge direction is vertical, determining the context model that comprises a highest probability assigned to a vertical filter shape;
when the edge direction is horizontal, determining the context model that comprises a highest probability assigned to a horizontal filter shape; or
when the edge direction is diagonal, determining the context model that comprises a highest probability assigned to a diagonal filter shape.
20. A method of processing visual media data, the method comprising:
processing a bitstream that includes the visual media data according to a format rule, wherein
the bitstream carries a plurality of pictures; and
the format rule specifies that:
coded information associated with one or more blocks in a current picture indicates that a cross-component sample offset (CCSO) filter is used for coding the one or more blocks, the CCSO filter being associated with a plurality of CCSO filter parameters;
one or more features associated with the one or more blocks are determined, the one or more features being used as a context for signaling a value associated with the CCSO filter;
a context model for signaling the value of at least a parameter of the CCSO filter in the plurality of CCSO filter parameters is determined based on the one or more features, the context model including different probability values associated with candidate values of the parameter of the CCSO filter with the one or more features being used as the context;
at least a signaled value associated with the CCSO filter is determined based on the context model; and
samples in the one or more blocks are reconstructed based on at least the signaled value associated with the CCSO filter.